Roles, strengths and challenges of using robots in interventions

for children with autism spectrum disorder (ASD)

Abstract

The aim of this research was to study roles, strengths and challenges of robot-mediated

interventions using robot KASPAR for children with autism spectrum disorder (ASD).

Twelve focus group sessions were organized in which 70 ASD care and/or education

professionals participated. Six roles for KASPAR were identified: provoker, reinforcer,

trainer, mediator, prompter, and diagnostic information provider. Strengths of KASPAR are

related to personalisation possibilities, its playfulness, the action - reaction principle, its

neutral expression, consistent and repetitive application of actions, possibilities to vary

behaviour in a controlled manner and having an extra hand. Challenges of working with

KASPAR were: limited reaction possibilities, possibility of children being scared of

KASPAR, difficulties with generalisation or transfer and finally potential dependence on

KASPAR.

Keywords: Autism spectrum disorders, Robot, KASPAR, ASD, Children, Robot Mediated

Intervention

- 1 -

Introduction

An upcoming intervention area with promising potential for children with autism spectrum

disorder (ASD) (American Psychiatric Association, 2013)) is robot-assisted therapy (RAT) or

robot-mediated intervention (RMI). A number of recent systematic literature reviews have

covered the majority of the work related to the use of robots for children with autism

(Begum, Serna, & Yanco, 2016; Cabibihan, Javed, Ang, & Aljunied, 2013; Diehl, Schmitt,

Villano, & Crowell, 2012; Huijnen, Lexis, Jansens, & de Witte, 2016; Pennisi, Tonacci,

Tartarisco, Billeci, Ruta, Gangemi, & Pioggia, 2016; Scassellati, Henny Admoni, & Matarić,

2012). However, despite a growing number of human-robot-interaction (HRI) studies for

children with autism of which the majority present promising results, advances in robot-

mediated interventions have only made minimal progress towards clinical applicability

(Begum et al., 2016). Begum et al. (2016) summarise a number of important reasons for this;

the potential end-users of this technology (i.e. people with ASD, their caregivers, clinicians)

are neither aware nor convinced of the role of robots in an ASD intervention (Diehl et al.,

2014, 2012); many of the robot-based studies for individuals with autism fail to meet criteria

commonly used to assess the outcome of an ASD intervention (Reed, Hyman, & Hirst, (2011;

Grynszpan, Weiss, Perez-Diaz, & Gal, 2014), and finally they argue that demonstrating the

‘likeability’ of robots for children with autism is “not sufficient to formally allow a robot to

co-locate and interact with a protected population such as individuals with autism”.

When studying robots presented literature, six different roles can be distinguished:

“diagnostic agent”, “friendly playmate”, “behaviour eliciting agent”, “social mediator”,

“social actor”, and “personal therapist” (Cabibihan et al., 2013). The authors came up with

these roles based on studies in the literature using robots for this target group. Apparently, as

Begum et al. (2016) argue, ASD professionals in practice seem to be unaware of the state of

- 2 -

the art as well as the possibilities of robots for this target group. At the same time, robot

developers are not always aware of the needs and capacities of children with ASD and those

who care for them. This creates a gap that hinders the creation of relevant and effective robot

interventions for this target group.

Pennisi et al. (2016) found positive implications in their review that robots might provide

therapists and teachers with new means to connect with people with autism in support

sessions. They report that participants with ASD showed social behaviour towards robots,

showed reduced repetitive and stereotyped behaviours and reported spontaneous use of

language in sessions with a robot. In some cases people with ASD show behaviours towards

robots that people without autism have towards humans (Pennisi et al., 2016). To benefit

from the positive effects of the use of a robot in interventions, higher levels of stimulation

(interaction) by the robot are better than lower levels of stimulation. However, despite these

positive findings, studies need to better clarify whether gender, intelligence and age of the

participants affect the outcome of the intervention and whether any beneficial effects are

apparent only during the session or whether they are transferred outside the session as well.

One of the robot platforms that is being applied in research for children with autism is robot

KASPAR (see Fig. 1 ) (Dautenhahn et al., 2009; Robins & Dautenhahn, 2014; Wainer,

Dautenhahn, Robins, & Amirabdollahian, 2010; Wainer, Robins, Amirabdollahian, &

Dautenhahn, 2014). KASPAR stands for “Kinesics and Synchronization in Personal Assistant

Robotics”, and is a semi-autonomous humanoid robot with the size of a sitting infant or

toddler and is minimally-expressive (Wainer, 2012).

[Place figure 1 here please]

KASPAR is developed at the Adaptive Systems Research Group at the University of

Hertfordshire (UK). It uses head, torso, arms and hand movements to make gestures and

- 3 -

(limited) facial expressions. Moreover, sound and speech can be used as additional

interaction mechanisms. Each arm has 3 degrees of freedom and different parts of its face

(e.g. eyes, mouth, eye lids) have separate motors that can be activated. KASPAR sits on a

table and cannot stand up or walk away (its legs are not actuated) (Wainer, 2012). KASPAR

can be controlled by activating its sensors in various body parts (hands, belly, feet, head) and

by using a pre-programmed remote control. Using the remote control, professionals can

initiate actions of KASPAR using the Wizard of Oz technique (in which the intention is that

the child does not notice that the robot is controlled by the professional). New KASPAR

behaviour, also called scenarios, can be easily created by using the KASPAR configuration

program installed on a laptop.

Research has shown that the use of KASPAR in interventions contributes to learning body

awareness, encouraging collaborative skills among children with autism (Wainer et al.,

2014); prolonging children’s attention span (Costa, Lehmann, Robins, Dautenhahn, & Soares,

2013); mediating and encouraging children’s social interaction with other people (Robins,

Dautenhahn, & Dickerson, 2009); and learning about body parts and appropriate physical

interaction (Costa, Lehmann, Dautenhahn, Robins, & Soares, 2015).

Reviews summarise that often the focus of HRI studies is on skills and behaviours relating to

social and communication impairments, but that there are also other relevant target areas in

which RMIs are expected to contribute to in a meaningful manner and to be of social

significance (Begum et al., 2016; Huijnen, Lexis, Jansens, et al., 2016). This question was

addressed in a study that intensely involved ASD professionals to better understand to which

therapy and educational objectives of children with ASD, robot KASPAR is expected to be

able to contribute to (Huijnen, Lexis, & de Witte, 2016). Indeed experts expect most potential

for KASPAR in the domains of “communication”, “social interaction and relations”, but also

- 4 -

for “play”. Next to this, results indicate that professionals expect KASPAR to have potential

for contributing to objectives in domains such as “emotional wellbeing” and “preschool

skills”. They particularly see a role for KASPAR in the following top 10 objectives:

“imitation in play”, “making contact”, “imitation in social/interpersonal interaction”, “turn-

taking behaviour”, “orientation to listen”, “social routines”, “attention”, “learning a new form

of communication”, “talk – use verbal abilities”, “train or practice skills”, “follow up

instructions”, “pose a question/ask for help”, “having fun, experiencing pleasure”, and

“developing interest in play” (Huijnen, Lexis, & de Witte, 2016). These insights on top 10

objectives provide understanding on the objectives that can be targeted by RMI, but not in

what manner and hów best to utilize the advantages of robots to complement existing

interventions already used in practice.

By intensively involving professionals such as therapists and special needs teachers in the

present study, we aim to minimize factors that hinder (clinical relevance and uptake and

increase chances for) clinical applicability. As mentioned, currently often professionals

working in practice with children with ASD are not aware nor convinced of the role of robots

in an ASD intervention. An important element in this is understanding what the potential may

be of robots in interventions for children with ASD according to these professionals and in

what manner both strengths as well as challenges can be taken into consideration when

developing interventions.

The aim of the current study was to gain insight into the potential of the robot KASPAR as

contribution to interventions according to practitioners in the field. In short, this study

addresses the following research questions:

- 5 -

What possible roles for KASPAR in an intervention for children with ASD are

suggested by professionals, and what strengths and challenges related to KASPAR do

they foresee?

Method

A qualitative study was performed which consisted of focus group sessions (Morgan &

Spanish, 1984) with professionals working in the ASD field. The term “professionals” is used

to denote practitioners who’s profession it is to work with children with autism in care and/or

special education. The aim was to gather insights on relevant aspects with respect to the role

KASPAR could have in an ASD intervention and to identify KASPAR’s strengths and

challenges related to using KASPAR in practice.

- 6 -

Setting and Participants

A number of organisations in the domain of ASD, in the south of the Netherlands, were

approached by the main researcher (CH). If they expressed their interest and willingness to

contribute to the study, a contact person from the organisations checked internally if there

were professionals that met the inclusion criteria of the study. Additionally, sampling was

conducted based on the snowball method keeping in mind a number of inclusion criteria for

participating respondents:

The professional works with children with ASD in their daily practice

The professional is working at a special school, care organisation or medical day

care centre targeted at children with ASD

Mastering the Dutch language

Being able to participate in a focus group session during the period of the study.

Procedure

- 7 -

The contact person at each organisation approached colleagues who met the inclusion criteria

to invite them for the focus group sessions. The sessions were organised at the premises of

the participating organisations. The duration of each session was about 2 hours. Two

researchers involved in the study were present during each session; one had the role of

session moderator and the other of note taker, observer and transcriber of the sessions

afterwards. Before the start of the session, informed consent papers (for making audio

recordings and pictures), participant demographic sheets and pens where distributed for each

participant. The focus groups started with a short welcome, an explanation of the (goal and

nature of the) session and a demonstration of KASPAR to give the participants a better idea

of what KASPAR is and what it can do. During the demo KASPAR introduced itself, played

a song, and additionally the possibilities were explained and shown (such as the use of

sensors, the freedom of movement of the motors, speech, sounds, remote control, as well as

the possibility to create personalised new scenarios and the use of additional attributes).

Participants were informed that behaviour of KASPAR needs to be programmed in advance

and that the operator who is remotely controlling KASPAR during a session, can select a

behaviour with or without a corresponding sound or utterance. After the introduction,

possible questions were answered and the actual focus group discussion started. The focus

group sessions were consistent in terms of structure and main questions that were asked.

Examples of questions are: do you see possibilities for using robot KASPAR in your

organisation? If so, for what objectives and in what way? Are there any challenges to take

into account? We stressed the importance of participants being open and free in their

reactions (e.g. that there is no “right” or “wrong” answer or opinion) and that every

contribution that people give is valuable.

Setting and Participants

- 8 -

A total number of 70 professionals participated during the focus group sessions; 60 were

female and 10 were male. All of them were recruited based on their expertise and experience

in working with children with ASD (e.g. teaching, providing training or care). The average

working experience for the professionals was 13.7 years with a standard deviation of 9.4

years. Table 1 summarizes the main characteristics of the participants. All the participants

had a positive attitude towards the use of technology in their daily practices. However, none

of them had previous knowledge of, or experience in, working with robots for children with

autism.

[Place table 1 here please]

In total, 22 different organisations were represented by 70 ASD professionals in 12 focus

group sessions. For people working for the same organisation the session was organised at

their premises (all in The Netherlands). The types of organisations that were represented

were: special needs schools, child and youth care organisations, (pedagogical) treatment

centres, and medical day care centres. The background of the participants varied; ranging

from ASD teachers and assistants at special needs schools, psychologists, speech therapists,

occupational therapist, physiotherapist, psychomotor therapist, behavioural therapists,

treatment coordinator, to people working in care management.

Data collection

- 9 -

The participant demographic forms were filled in and collected on paper and the information

was stored in an excel sheet. Audio recordings were made during the 12 focus group sessions

and verbatim transcripts were written in Word. Nvivo was used to analyse the verbal

material.

Data analysis

To analyse and interpret the content of the qualitative data of the focus groups, conventional

content analysis was applied (Hsieh & Shannon, 2005). This entailed that coding categories

were derived directly in an inductive manner from the text data rather than from an existing

predefined coding scheme. Data triangulation was used to ensure data integrity and validity.

Two persons were involved in the collection and analysis of the data (investigator

triangulation) (Carter, Bryant-Lukosius, DiCenso, Blythe, & Neville, 2014)). Two

researchers were independently involved in creating the coding scheme based on the

transcripts from the sessions to ensure inter-subjectivity. Both were present during the focus

group sessions. In other to reach saturation, two analytical sessions were organised. When all

the focus group sessions were transcribed by one researcher, both researchers read two

sessions with the goal to identify main themes or codes to create a tentative coding scheme.

After this, an analytical session was organised to compare, discuss and align these two coding

schemes. The codes were discussed to understand how the data was perceived by both

researchers. On a semantic level (the meaning of the data) the coding schemes were very

similar, and the best wording (label) for each code was agreed upon. The resulting coding

scheme was used by the main researcher to code three more focus group sessions.

Subsequently, an additional analytical session with the two researchers was scheduled to

finalise the coding scheme to ensure saturation. During the second analytical session some

codes were slightly rephrased and finally coding consensus and information saturation was

achieved. There was agreement on what code belonged to a certain piece of transcription. The

- 10 -

rest of the focus group sessions were then coded by the main researcher using this final

coding scheme (see the appendix for the final coding scheme).

- 11 -

Results

Results of the focus groups indicated that six different roles can be distinguished for

KASPAR based on the input of the professionals. They argue that depending on the needs

and capabilities of the professionals and the training or educational objectives for a child at a

certain moment in time, as well as the available other interventions, a certain role for

KASPAR may be preferred and chosen. Moreover, rich insight was gathered into the

strengths of KASPAR as well as the challenges related to creating robot mediated

interventions.

Possible roles for KASPAR

Below, the six roles that professionals envisioned for KASPAR are presented. Professionals

indicated that these roles are not mutually exclusive; it may happen that two or more roles

apply for a certain task and that another task requires another role. They indicated that the

professional shall choose what role(s) are important to reach a certain goal.

Provoker

KASPAR is expected to function as a kind of magnet to (some) children; one that evokes,

provokes, elicits, triggers, and stimulates them to engage in interaction. This role was

suggested in 6 (of the 12) sessions. KASPAR may remove some barriers that normally

prevent these children from feeling free and behaving in certain desired ways. A speech

therapist at a special needs school indicated “KASPAR is calm, and manageable for children,

he can provoke the children at a safe, calm and step by step manner” (Focus Group (FG) 6,

speech therapist 1, special needs school, Female (F), 16 years of working experience with

children with ASD). A special needs teacher mentioned “KASPAR might be able to trigger

children and give another way of motivation and attention.” (FG4, special needs teacher,

- 12 -

special needs school, F, 7). Once KASPAR has won the children’s attention, “he” can start

“working” with them. Professionals highlighted that, currently, often the interaction between

a professional and a child is unidirectional; it starts from the professional and stops there

when the child does not respond. KASPAR, in their views, may be able to function as a

trigger or stimulus that causes initiative from the child. Because KASPAR can always give

(the same) reaction, the child might seek contact him-/herself. This creates a safe and

predictable atmosphere for the child. KASPAR may stimulate or remove an obstacle so that

children might start doing things they normally would not do. A special needs teacher

(specialised in physical exercise) indicated that “KASPAR might be able to give a trigger so

that the child takes initiative. Because KASPAR will always give a reaction, the child might

seek contact.” (FG11, special needs teacher, special needs school, Male (M), 17).

Reinforcer

Secondly, professionals indicated that KASPAR can be used to generate success experiences,

to positively reinforce the children by giving compliments. “KASPAR can give compliments

to the children” (FG9, pedagogical staff, ASD care centre, F, 8). This reinforcer role was

suggested in 7 (of the 12) sessions. If needed, the professional can control the robot to deliver

a certain reaction. Appropriate behaviour of the child can immediately be rewarded by a

(consistent) compliment of the robot, initiated by the professional. In the beginning one may

give a positive reinforcement after each instance of appropriate behaviour, and over time this

can be given less frequently when the child learns. “We are able to control KASPAR. The

reinforcement can be given immediately, or a bit later, or even later after the action. You can

give a positive reinforcement after 1 correct action, later after two correct actions...you can

play with it” (FG12, special needs teacher, special needs school, F, 14). When a child finishes

a task, KASPAR might say “Well done [name of the child], you finished your task”. In this

- 13 -

way the child experiences moments of joy, and success which are very important for them

since often things are difficult, challenging or even impossible for them. If the child does

something which is less positive, KASPAR might say in a neutral voice without emotions

“try it again”. Professionals stressed that it is crucial that children experience success

experiences, KASPAR could contribute to these positive moments. “He [child] will always

receive the proper reaction. So when he takes initiative, this will be rewarded by KASPAR as

an incentive” (FG5, ASD teacher 1, special needs day care centre, F, 17).

Trainer

Thirdly, professionals suggested that KASPAR can be used to repeat certain actions over and

over again so that the child can train and practice a (new) skill. “KASPAR could take over a

trainer function. I can say it is your turn, or KASPAR says it. I am curious to see if they learn

faster if a robot tells it than I” (FG9, trainer, centre for ASD care, F, 4). This role has been

suggested in 9 of the 12 sessions. This can be related to a wide and diverse scope of skills,

examples are given of imitation skills, following up instructions, learning to cope with

proximity, learning to work (independently), understanding appropriate behaviour in different

contexts, and solution-oriented behaviour. “Especially with imitation this seems handy;

KASPAR can give the example, you can tell the child to look at KASPAR to see how it should

be done. Normally I give the examples myself plus I have to instruct the child at the same

time” (FG12, speech therapist, medical day care centre, F, 20).

Moreover, a number of special needs teachers suggested that realistic examples from the

child’s life could be taken to practice with KASPAR in the sessions. Parents could, for

example, provide situations of things that occurred, which will be worked with in the sessions

so that the child can actually learn from it. In this case, KASPAR can be used in many

different situations that make sense in daily life. By explaining for example that proper

- 14 -

behaviour can be context dependent. A certain behaviour might mean different things in

different contexts. For example, by simply changing KASPAR’s clothes into a police

uniform, the gesture of holding up one hand would change from waving (what it would be

when you see a friend on the street) into a formal stop sign.

Mediator

Additionally, professionals indicated that KASPAR could be used as a medium to enable

contact, interaction and communication with the child and another person. This role was

mentioned in 8 of the 12 sessions and participants had lively discussions about it. They

envisioned that a triangle of joint attention could be established via KASPAR (to a

professional, another child or an object). Joint attention is a prerequisite to learn to interact

together and KASPAR might help in establishing it. At first, KASPAR has the focus of

attention which can be extended to other objects or topics. He can be an attention-,

communication- and interaction mediator. Also KASPAR does not always have to be directly

or primarily used to actually train certain skills, but he can function as a medium to engage in

a moment or activity together as well. It can be a kind of subject of conversation that

functions as a bridge to other topics; to support triangulation (involving a person and another

person and an object or the robot). “I can imagine this happening with two of those little ones

of us. Play a turn taking game together, listen to each other, wait for each other. KASPAR

could be the third person taking the lead, for example “hey guys…” (FG9, ASD trainer 1,

centre for ASD care, F, 4). Professionals envisioned that it is not always so important what

KASPAR does, but that just simply being there would already make a difference. “For me as

ASD therapist, there are so many means, but sometimes I cannot find a fit to the needs of the

child. We could use KASPAR as another medium. It will not be so much about what he does,

but more about that he is there. Then you use KASPAR as a medium for shared activity, or as

- 15 -

a conversation topic to reach other topics, together” (FG9, ASD trainer 2, centre for ASD

care, F, 8). Moreover, KASPAR could be used to advise the children to talk to the teacher,

their mum or dad. Finally, they indicated that children could also use KASPAR as a voice of

their own, to express their wishes or needs if they experience difficulties with this themselves

(similar as they do with a speech computer in a sense). An ASD care therapist who supports

children and parents with ASD also a mum of two sons with ASD argued (FG7, ASD care

therapist, care organisation, F, 18): “It would be wonderful if we could use KASPAR to

stimulate or learn solution oriented behaviour. Imagine I am angry, that you can use

KASPAR to show what being angry looks like. And that together you can start to think, what

would KASPAR do now? Because that is not as scary as when you have to look at yourself.

That would be wonderful, if you can use KASPAR as a tool to discuss feelings…that seems

less intimidating to me. It could work with other feelings as well. Some children are too

happy, what is appropriate and when. KASPAR can mediate in that.”

Prompter

A fifth role that was mentioned was that KASPAR might be used to give (concrete, specific

and clear unambiguous) instructions or prompts to the children. This role was suggested in 8

of the 12 sessions and participants considered this role of value in practice. They indicated

that often children need quite a high level of (physical) proximity to complete a task (on their

own). They know, for example, how to perform a certain task, but as soon as they finished

that task, they wait and wait. Often they need a person to actually tell them that they need to

move on to the next task (which in turn they can perfectly complete themselves again).

KASPAR might fulfil that prompting role to give them a little ‘hand’ to pursue their work. At

the end of their task, a KASPAR picto could be used, so that they know they have to interact

with KASPAR, who might say “Well done [name child] now you can take the yellow basket

- 16 -

and make the next task”. In the words of a teacher assistant (FG6, teacher assistant, special

needs school, F, 11): “KASPAR can give instructions to the child, go on, well done, please sit

down, take the next task, please continue”. Or more in general, KASPAR can be used to give

instructions or a day/task structure to the children. One example was given by a teacher of a

special education school (FG4, special needs teacher, special needs school, F, 25): “When the

children arrive at school in the morning, they need to do three things: hang up their coat, put

their breadbox in the drawer, and sit at their desk. These are three different things and

KASPAR could provide a structure and guidance to complete these step.” Or as another

participant indicated “KASPAR might be used to plan and organise when there is a range of

activities….. “in 5 minutes we will stop and have lunch” (FG8, ASD coach, care

organization, M, 20). In this case the dependency of the human might become smaller, and

step by step one can work towards a more independent working style (if the dependency of

KASPAR of course is gradually decreased as well). If a certain child has sensory difficulties

when there are more children around, KASPAR could be placed next to him/her so that more

isolated instructions or prompts can be given.

Diagnostic information provider

A sixth and final role that was suggested by the professionals was a diagnostic information

provider. By observing how the child interacts with KASPAR, professionals might learn

more about the behaviour of the child. This role was brought up by participants in 4 of the 12

sessions. It might provide some diagnostic information that would be difficult to get when the

child was interacting with the professional rather than with the robot. From the perspective of

a person without autism, KASPAR might look ‘clean’, awkward, or even ‘scary’ due to the

absence facial expressions and details. For people without autism, facial expressions often

provide a sense of safety, security, clarity. For children with autism, this is (completely)

- 17 -

different, which might be a reason why they show different reactions and behaviour to

KASPAR than to a human being. Professionals expected to be surprised by the children’s

reactions to KASPAR. This might help professionals to understand the children better and

possibly learn them more about their own behaviour towards the children (i.e. “why does the

child react in this way to KASPAR and differently to me”) and give them more insight in

how to improve their behaviour towards the children. So, professionals thought that from

observing the child’s interaction with KASPAR, professionals might learn about the child on

the one hand, and about their own behaviour on the other hand. A number of quotes from

participants illustrate this: “KASPAR can help me understand why does this child react this

way and another child that way. And what does the child apparently need, also from me. In

this way it is a beautiful diagnostic information tool” (FG7, mental health psychologist,

ASD care organisation, F, 30); “I do think it works in two directions; it is not just very clear

for the child but we [teachers] we will see that KASPAR will provoke reactions that we just

cannot provoke. Then you will start to think “why does he [child] do it with KASPAR and not

with me? Maybe in this way you will search yourself for better ways of communicating with

the child” (FG, special needs teacher, ); and “KASPAR can help us to get a better

understanding in what a child can actually do” (FG5, ASD teacher 1, special needs day care

centre, F, 17).

- 18 -

Strengths of KASPAR

The focus group results indicated that professionals envisioned a number of strengths

inherent to KASPAR that could be exploited or utilised in order to optimally complement

current ASD practices. Often these strengths are expected to create desirable states of

wellbeing for the child such as safety and trust which better enable the child to feel pleasant,

motivated and more able to develop to their fullest potential. These characteristics were the

following:

1. Personalize and apply in a multifunctional manner; KASPAR can be programmed

in a tailored individualised manner to meet every child’s needs, wishes, learning

objectives, capabilities and preferences. Personalisation is important for children with

autism and KASPAR is expected to be able to have an answer to this since new

scenarios and KASPAR behaviour can easily be created or fine-tuned. This applied

both for the training and educational goals that are relevant for that particular child

and will be addressed using the robot, but also to the way these are addressed.

KASPAR should look accessible and friendly and children should be able to identify

with him. KASPAR is currently a boy, but can be changed into a girl when his hair,

clothes and voice is adapted into a female version of the robot.

2. Playful; it is expected that (some) children will enjoy interacting with KASPAR, that

they will like it, experience fun, which increases chances for enjoyment, motivation,

attention and drive for initiative or exploration. For children with autism this can be a

very important positive aspect.

3. Action - reaction; the child is in control; the initiative lies in the hands of the child.

Many children with autism are fond of interactive technology such as computers or

tablets. One of the underlying causes is, in the eyes of the professional, that the child

- 19 -

is in charge of the initiative and control which gives a feeling of predictability and

safety. It is expected that the ‘action-reaction’ mechanism of KASPAR is a strong

asset for being a successful tool for children with autism.

4. Neutrality; approachability, lack of “noise” or ambiguous disturbances; due to

KASPAR’s lack of extreme realistic human facial expressions and emotions he is

expected to be more approachable (than humans) for children with autism. It is

expected, that in the eyes of the children, KASPAR is thought to be more ‘clean’,

predictable, safe, less distracting, trustable, and less ambiguous and less threatening

than a human person would be.

5. Consistent and repetitive application (creates predictability); KASPAR is able to

say or do something in exact the same manner for an endless number of times. For

humans this is not possible: we will always (unconsciously and unintentionally) vary

in some way; our tone of voice, volume, pitch, facial expressions, or speed might

alter. For training and practicing purposes this consistent repetition is expected to be

extremely beneficial. KASPAR’s behaviour and/or reaction can be the same, over and

over again. This consistency creates predictability which in turn can create a feeling

of safety for the children (i.e. KASPAR will not perform any unexpected ‘strange’

actions). KASPAR always sits, so they do not have to be ‘afraid’ that he will just get

up and move away. Moreover, KASPAR is always there and available, whereas some

professionals work part-time, are replaced or are on (sick)leave every now and then

which might upset some children. KASPAR might be a stable and constant factor

alongside these (changing) professionals.

6. Vary in a controlled manner; according to professionals, with KASPAR you have

the ability to deliberately chance only some selective and isolated aspects in his

- 20 -

reaction or behaviour. Variation or change can be done in a controlled and conscious

manner. For humans this might be more challenging since we unconsciously and

unintentionally sent numerous messages with our voice, facial expressions, odours,

body posture that children sense, perceive and possibly react to. KASPAR is in that

sense more focused and selectively controllable when professionals wish to change or

vary gradually to work towards transfer and generalisation.

7. Extra hand; professionals highlighted that at the moment often professionals occupy

a double or triple function. They fulfil multiple roles simultaneously often at a single

moment in time in one person; a trainer, an observer, a motivator, a corrector and

often a kind of coach. Professionals indicated that this causes a large burden on them

and might cause a lack of clarity or maybe even distress or overload for the child.

With the use of KASPAR in one of these roles, the professional gets more ‘space’ to

focus and or even to take some (physical) distance. For the child it is expected to be

less ambiguous.

Challenges related to KASPAR

The results also indicated that professionals envisioned a number of challenges related

KASPAR that should be given attention to in order to optimally complement the current ASD

practices. These challenges were the following:

1. Limited reaction possibilities; according to the professionals, currently, KASPAR

has limited mobility capabilities (i.e. he cannot walk, grasp or fetch objects, make fine

gestures with his hands or fingers). Since KASPAR is semi-autonomous, a number of

pre-defined actions can be programed on the remote control. This means the

professional has these and only these reaction possibilities at hand; KASPAR itself

has no contextual sensitivity to purposefully react in a situation.

- 21 -

2. Some children might be scared of KASPAR; professionals thought that it might

very well be that not all children like KASPAR, some children might even experience

aversion towards the robot.

3. Generalisation / transfer; professionals indicated that KASPAR might be able to

train new skills, however, the intention is not to teach the child to optimally and only

perfectly behave in the interaction with KASPAR. The goal is that children are able to

apply the learned skills also in real live situations. Professionals raised the issue of

transfer or generalisation to humans and/or other situations. Also, some professionals

were afraid that some children might copy KASPAR in an extreme realistic manner

causing them to show ‘robotic’ behaviour themselves.

4. Dependence on KASPAR; it may occur that children with autism have a dependency

to some of their teachers or therapists. If a child learns some words with KASPAR as

a trainer, coach, or medium in the intervention, professionals wondered if this

dependency might shift towards KASPAR. Therefore, attention needs to be devoted

to the matter of dependency as well.

- 22 -

Discussion and Conclusion

Results from focus groups with professionals working intensively with children with autism

and other stakeholders delivered 6 roles for a robot in the work they do: provoker, reinforcer,

trainer, mediator, prompter, and a diagnostic information provider. These roles were defined

based on their expertise of what children with ASD need in education and training in order to

support their independence in daily life. The roles of the robot were formulated in terms of

delivering added value to the already existing work practices and interventions.

Professionals considered the use of robots to be of high potential. In order to achieve this

potential, it is crucial that the strong assets of robots (e.g. action-reaction principle, consistent

and repetitive application) are carefully implemented in practice while taking into account

and utilising the strong characteristics of people (e.g. being able to ‘read’ the child with

autism very well). As a result, they expect that robots can become valuable tools in the hands

of professionals.

Results indicated that, depending on the needs of the child and the professional a certain role

might be chosen in a certain moment while another role might be needed at another moment

in time. Moreover, findings showed that these roles are not mutually exclusive, multiple roles

can be applicable or needed for a certain task or activity. Professionals indicated, for

example, that for the role of trainer compliments of the reinforcer role are needed as well.

The roles from the ASD professionals are quite similar to those identified in state of the art

robotics: “behaviour eliciting agent”, “diagnostic agent”, “friendly playmate”, “social

mediator”, “social actor”, and “personal therapist” (Cabibihan et al., 2013). On the other

hand, when taking a closer look, the roles from the professionals seemed to be more concrete

and slightly more intervention oriented when implementing them in a real education or care

practice. For example, when looking at the “diagnostic information provider”; the

- 23 -

professionals envision a two way interaction, not only to establish information related to the

diagnoses of the child’s abilities (as suggested in the categorization based on the state of the

art of robot literature), but also a feedback mechanism for the professionals to learn and

possibly adjust their own behaviour towards the child.

The goal of the study was to get insight into as many different roles that could be envisioned

for robot KASPAR in ASD interventions. Since people with autism might have a different

perspective or experience, we also invited three adults with ASD in an additional focus group

as an extra source of information and to check if their ideas align with what the professionals

mentioned. They indeed confirmed the results of the professionals. Moreover, they delivered

interesting additional information. They came up with a 7th possible role; a buddy. They

argued that KASPAR might say “Hello, my name is KASPAR, I am a little bit different, just

like you”. So that the child learns it is alright to be different and still be part of the group.

This might enable a different connection or relationship, more like a friend rather than a

teacher. An adult male with ASD indicated “He [KASPAR] is a friend and does not have a

teaching role”. If the child then establishes a trust relationship with KASPAR, it might feel

safer. KASPAR might be able to ‘help’ then in difficult cases in which it is hard for the child

to express its needs or wishes to a human. KASPAR might fulfil the role to stand next to the

child and help to understand what really happened.

More research is needed to systematically and intensively involve people with ASD in the

process of researching and developing valuable robot-mediated interventions for children

with ASD.

A strength of the present study was intensively involving professionals in the creation of new

robot mediated interventions to utility mutual awareness and expertise to better guide robot

mediated interventions. Professionals see a strength and advantages of robots for this target

- 24 -

group, maybe even more than in other domains were robots are suggested for care (e.g.

independent living, service robotics for elderly care as described in for example in (Draper et

al., 2014)). Possibly this is due to the nature and complexity of the disorder, the difficulty of

delivering proper care especially as human beings and the open mind that many of these

professionals have towards trying out new ways of working with this target group. More

robotic appearance and behaviour might be ‘better’ or preferred in some moments with this

target group which is particularly difficult for expressive people. This relates to findings in

which children showed more interaction with a ‘robot looking man’ than with a typical

human being (Robins, Dautenhahn, & Dubowski, 2004).

The potential that was identified using robot KASPAR, can also be applied to many similar

other robots that are used in interventions for training or education for children with ASD, as

often on an abstract level many robots have similar characteristics. For example, other robots

(e.g. Nao as presented in (Miskam et al., 2013; Tapus et al., 2012) can also provide a

possibility to personalise the behaviour of a robot to the needs of a specific child, can be

playful in the interaction or appearance, and provide a kind of action – reaction mechanism

that allows the child to be in control. Depending on the appearance of the robot platform,

interaction with a robot can be neutral and approachable for a child with autism. Depending

on the implementation (or to some extent the level of autonomy) of the robot, many robots

can provide consistent interaction, the option to vary in a controlled manner and provide an

extra hand for therapists or teachers. The same holds for the need to pay attention to a number

of aspects when working with robots for children with autism; the robot will always have

some limitations, for some children or some moments a robot will not be the most optimal

choice, prevent dependence on the robot, and take into account matters of generalisation and

transfer. The robot shall be an additional tool in the hands of the professionals, not a goal in

- 25 -

itself. The aim is to support children in their development so that they can function in the real

world with real people, not (only) show great performance in front of a robot.

A limitation of the study could be the snowballing sampling method that was used; due to

possible anchoring (not knowing if the sample is an accurate reading of the target

population). However, this is a high risk only if only a few people would be included, which

is not the case in this study in which a large number diverse professionals (N=70) participated

from many different and diverse organisations.

Previous studies already show that professionals have positive expectations of using

KASPAR to work on the education and/or training objectives for children with autism; that

they see potential for KASPAR for more ASD training and educational objectives than

currently proposed in robot studies in peer reviewed literature (Huijnen, Lexis, & de Witte,

2016; Huijnen, Lexis, Jansens, et al., 2016). They clearly can see KASPAR as a tool in the

hands of the professional to work on a larger range of objectives. This study contributes to

this mission by creating understanding of where the strengths or advantages of robots might

be in an intervention for children with autism. By intensively involving professionals we

aimed to increase chances for clinical relevance and uptake and overcome typical barriers for

robot mediated interventions to reach clinical applicability as stated by Begum et al. (2016)

(such as unawareness or lack of believe in the potential of robots by end-users).

The next step will be to co-create and pilot new robot-mediated interventions using these

insights as well as the gained awareness and knowledge of the professionals in the field as

well as people with autism. Moreover, effort will be given to minimize the reported

challenges, for example by co-creating new robot mediated interventions aiming to enhance

the robot design and optimize personalised interaction possibilities and dialogues. When truly

implementing robots into current practices, also aspects other than the roles, strengths and

- 26 -

challenges need to be in place. These can be technical aspects such as technical infrastructure,

maintenance effort and costs, as well as stability. Additionally, also many practical

requirements need to be taken into account in order to truly implement robot assisted

interventions (e.g. a specification of the target group, factors related to the environment,

integration into common practices and individualised education/training plans) (Huijnen,

Lexis, Jansens, & Witte, 2017).

Ethical approval: This article does not contain any studies with human participants performed

by any of the authors.

Informed consent: informed consent was obtained from all the individual participants

included in the study.

- 27 -

References

American Psychiatric Association. (2013). Diagnostic and Statistical Manual of Mental

Disorders (DSM-5®). Book, American Psychiatric Pub.

Begum, M., Serna, R. W., & Yanco, H. A. (2016). Are robots ready to deliver autism

interventions? a comprehensive review. International Journal of Social Robotics, 8(2),

157-181.

Cabibihan, J. J., Javed, H., Ang, M., & Aljunied, S. M. (2013). Why Robots? A Survey on

the Roles and Benefits of Social Robots in the Therapy of Children with Autism.

International Journal of Social Robotics, 5(4), 593–618. http://doi.org/10.1007/s12369-

013-0202-2

Carter, N., Bryant-Lukosius, D., DiCenso, A., Blythe, J., & Neville, A.J. (2014). The use of

triangulation in qualitative research. Oncology nursing forum, 41 5, 545-7.

Costa, S., Lehmann, H., Dautenhahn, K., Robins, B., & Soares, F. (2015). Using a humanoid

robot to elicit body awareness and appropriate physical interaction in children with

autism. International journal of social robotics, 7(2), 265-278.

Costa, S., Lehmann, H., Robins, B., Dautenhahn, K., & Soares, F. (2013). ” Where is your

nose?”: developing body awareness skills among children with autism using a humanoid

robot. In ACHI 2013, the Sixth International Conference on Advances in Computer-

Human Interactions (pp. 117-122). IARIA, 2013.

Dautenhahn, K., Nehaniv, C. L., Walters, M. L., Robins, B., Kose-Bagci, H., Mirza, N. A., &

Blow, M. (2009). KASPAR–a minimally expressive humanoid robot for human–robot

interaction research. Applied Bionics and Biomechanics, 6(3-4), 369-397.

- 28 -

Diehl, J. J., Crowell, C. R., Villano, M., Wier, K., Tang, K., & Riek, L. D. (2014). Clinical

Applications of Robots in Autism Spectrum Disorder Diagnosis and Treatment. In B. V.

Patel, R. V. Preedy, & R. C. Martin (Eds.), Comprehensive Guide to Autism (pp. 411–

422). New York, NY: Springer New York. http://doi.org/10.1007/978-1-4614-4788-

7_14

Diehl, J. J., Schmitt, L. M., Villano, M., & Crowell, C. R. (2012). The clinical use of robots

for individuals with autism spectrum disorders: A critical review. Research in autism

spectrum disorders, 6(1), 249-262.

Draper, H., Sorell, T., Bedaf, S., Syrdal, D. S., Gutierrez-Ruiz, C., Duclos, A., &

Amirabdollahian, F. (2014). Ethical dimensions of human-robot interactions in the care

of older people: Insights from 21 focus groups convened in the UK, France and the

Netherlands. In International Conference on Social Robotics (pp. 135-145). Springer,

Cham.

Grynszpan, O., Weiss, P. L., Perez-Diaz, F., & Gal, E. (2014). Innovative technology-based

interventions for autism spectrum disorders: a meta-analysis. Autism, 18(4), 346-361.

Hsieh, H. F., & Shannon, S. E. (2005). Three Approaches to Qualitative Content Analysis.

Qualitative Health Research, 15(9), 1277–1288.

Huijnen, C. A., Lexis, M. A., & de Witte, L. P. (2016). Matching robot KASPAR to autism

spectrum disorder (ASD) therapy and educational goals. International Journal of Social

Robotics, 8(4), 445-455.

Huijnen, C. A., Lexis, M. A., Jansens, R., & de Witte, L. P. (2016). Mapping robots to

therapy and educational objectives for children with autism spectrum disorder. Journal

of autism and developmental disorders, 46(6), 2100-2114.

- 29 -

Huijnen, C. A., Lexis, M. A., Jansens, R., & de Witte, L. P. (2017). How to Implement

Robots in Interventions for Children with Autism? A Co-creation Study Involving

People with Autism, Parents and Professionals. Journal of autism and developmental

disorders, 47(10), 3079-3096.

Miskam, M. A., Hamid, M. A. C., Yussof, H., Shamsuddin, S., Malik, N. A., & Basir, S. N.

(2013). Study on social interaction between children with autism and humanoid robot

NAO. In Applied Mechanics and Materials (Vol. 393, pp. 573-578). Trans Tech

Publications.

Morgan, D. L., & Spanish, M. T. (1984). Focus groups: A new tool for qualitative research.

Qualitative Sociology, 7(3), 253–270.

Pennisi, P., Tonacci, A., Tartarisco, G., Billeci, L., Ruta, L., Gangemi, S., & Pioggia, G.

(2016). Autism and social robotics: A systematic review. Autism Research, 9(2), 165-

183.

Reed, F. D. D., Hyman, S. R., & Hirst, J. M. (2011). Applications of technology to teach

social skills to children with autism. Research in Autism Spectrum Disorders, 5(3),

1003-1010.

Robins, B., & Dautenhahn, K. (2014). Tactile Interactions with a Humanoid Robot: Novel

Play Scenario Implementations with Children with Autism. International Journal of

Social Robotics, 6(3), 397–415. http://doi.org/10.1007/s12369-014-0228-0

Robins, B., Dautenhahn, K., & Dickerson, P. (2009). From isolation to communication: A

case study evaluation of robot assisted play for children with autism with a minimally

expressive humanoid robot. Proceedings of the 2nd International Conferences on

Advances in Computer-Human Interactions, ACHI 2009, 205–211.

- 30 -

Robins, B., Dautenhahn, K., & Dubowski, J. (2004). Investigating autistic children’s attitudes

towards strangers with the theatrical robot - a new experimental paradigm in human-

robot interaction studies. In RO-MAN 2004. 13th IEEE International Workshop on

Robot and Human Interactive Communication (IEEE Catalog No.04TH8759) (pp. 557–

562).

Scassellati, B., Henny Admoni, & Matarić, M. (2012). Robots for Use in Autism Research.

Annual Review of Biomedical Engineering, 14(1), 275–294.

http://doi.org/10.1146/annurev-bioeng-071811-150036

Tapus, A., Peca, A., Aly, A., Pop, C., Jisa, L., Pintea, S., ... & David, D. O. (2012). Children

with autism social engagement in interaction with Nao, an imitative robot: A series of

single case experiments. Interaction studies, 13(3), 315-347.

Wainer, J. (2012). Facilitating collaboration among children with autism through robot-

assisted play (Thesis). University of Hertfordshire.

Wainer, J., Dautenhahn, K., Robins, B., & Amirabdollahian, F. (2010). Collaborating with

Kaspar: Using an autonomous humanoid robot to foster cooperative dyadic play among

children with autism. In 2010 10th IEEE-RAS International Conference on Humanoid

Robots, Humanoids 2010 (pp. 631–638). http://doi.org/10.1109/ICHR.2010.5686346

Wainer, J., Robins, B., Amirabdollahian, F., & Dautenhahn, K. (2014). Using the humanoid

robot KASPAR to autonomously play triadic games and facilitate collaborative play

among children with autism. IEEE Transactions on Autonomous Mental Development,

6(3), 183–199. http://doi.org/10.1109/TAMD.2014.2303116

- 31 -

APPENDIX

This is the final coding scheme used to code the focus group sessions.

Possible roles for KASPAR

1. Provoker

2. Reinforcer

3. Trainer

4. Mediator

5. Prompter

6. Diagnostic information provider

7. Buddy

Why could KASPAR work - strengths of KASPAR

o Personalize and apply in multifunctional manner

o Vary in a controlled manner

o Action –reaction

o Playful

o Neutrality

o Consistent and repetitive application

o Extra hand

Why could KASPAR NOT work – challenges related to KASPAR

1. Generalisation / Transfer

- 32 -

2. Limited reaction possibilities

3. Some children might be scared of KASPAR

4. Dependence on KASPAR

KASPAR

Appearance

Voice or Sound

Operation of KASPAR

Behaviour / actions of KASPAR

KASPAR’s Attributes / Toolbox

Target Group

Specification of target who will probably benefit from KASPAR

Specification of target who will probably NOT benefit from KASPAR

Environment

Professional

Intervention implementation

Integration in individualised education/therapy plan

Phase in the intervention trajectory

Session characteristics

individual vs group, spontaneous vs structured, duration

integration on organisation level and connecting to parents

- 33 -

- 34 -

Fig. 1 Robot KASPAR

Table 1. Demographic characteristics of the participants

DESCRIPTION/VARIABLE VALUE (n (%))GENDER

Male n=10 (14%)Female n=60 (86%)

Total (N=70)

NUMBER OF YEARS WORKING EXPERIENCE WITH ASD (PROFESSIONALS)

0-5 years n=15 (21%)6-10 years n=19 (26%)11-15 years n=10 14%)16-20 years n=14 (19%)21-25 years n=3 (4%)26-30 years n=5 (6%)31-35 years n=4 (5%)

- 35 -